Motion Search Precision in TMPGEnc Plus

[aamir12345678]

I have a fairly, high-action anime that I'm encoding to Mpeg2 with TMPGEnc. I'm pretty much overshooting the required bitrate as is based on a CQ 87 encode test. It requires about a 4400 bitrate; I'm giving it a 4800 bitrate. I'm doing a 2-pass, using 1000-4800-8600 as min-avg-max. I'm using 9 Bit for the DC Comp Precision. What should I set the Motion Search Precision if I want to maintain max quality?

[Cobra]

For maximum quality, the slowest motion search is the one you should use. However, it has been found that the trade-off in speed is generally considered not worth the fractional gain in quality (personally, I can't even see the difference) to "High".

Looking at your system specification, I would be tempted to use "High (slower)".

Of course, it is a matter of personal choice - if you want the best possible video (even by a whisker) then go for the best and enjoy!

Cobra

[hiptune]

I have a fairly, high-action anime that I'm encoding to Mpeg2 with TMPGEnc. I'm pretty much overshooting the required bitrate as is based on a CQ 87 encode test. It requires about a 4400 bitrate; I'm giving it a 4800 bitrate. I'm doing a 2-pass, using 1000-4800-8600 as min-avg-max. I'm using 9 Bit for the DC Comp Precision. What should I set the Motion Search Precision if I want to maintain max quality?

If your program is shorter than 60 min. I personally would use CBR and mastering quality, and the bit rate that makes you happy.

Jeff

[NamPla]

I really have to recommend LordSmurf's TMPGEnc settings here:

http://www.digitalfaq.com/convert/tmpgenc/tmpgencplus.htm

Tho' dubious at first about sacrificing quality, after extensive testing I have to totally agree with him. The "sacrifice" is negligible, non-existent. Plus the added bonus of accelerated encoding times.

CQ-VBR
Motion Search Estimate (fast)
DC 8 or 9 (I use 9 for DVD, only because it doesn't add much to the encoding time over 8, which LordSmurf recommends. Using 10 just eats up unnecessary bitrate, or something).

This works superb for me, even with high-action/fast-movement scenes, explosions, giant monsters trashing cities, etc... Tho' I confess I haven't tried encoding anime...

TMPGEnc's Noise Reduction filter is terrific, but throw a Convolution3D into an avisynth script and it's even better - and MUCH MUCH faster - 24 odd hours down to 5 or so. Convolution3D has various anime-specific presets, too.

Well there's my 2 cents' worth. Cheers.[/i]

[fmctm1sw]

I've seen guides that recommend "motion search precision" for MPEG-2 video. It's fastest too.

*I don't think that's the name of it, it's something like that...

[Sillyname]

I'm doin' a concert video and it's got darkness(gain grain!!!), smoke, strobes, colored lights, video projections and some crazy fellas right in the middle of it all. It is busy...

I've been messing around with TMPG and actually found the lowest motion estimate search setting the best for the most difference between screens. Let me explain.

When you have one frame with a strobe flash on it and the next without, motion estimation will not do. If you have too much going on, plus MTV style cuts the computer can't estimate much of anything but cause more artifacts.

Automatic gain granularity on a camera can be especially crappy as the individual pixels sorta randomly boost areas of the shot.

[lordsmurf]

Estimate will be fine for animation (cartoons).

[aamir12345678]

--I moved down to High (Slower), and it made 2 hours what was taking 3 - 4. I don't notice any quality difference. The reason I haven't gone down to Motion Estimate is because I've seen some tests (forgot site unfortunately) that show the difference between search precisions at ample bitrates, and Motion Estimate produced more visible artifacts than High or Highest or even Normal. It might depend on what type of video you have, but I think it's my best bet to just stick with High.

--Thanks for all of your help.

[n3vino]

Using tmpgenc 2.5 to encode, I use 8000 bitrate and slow search high quality (not the slow highest quality).

If anyone has an opinion, which of these two parameters is more significant in quality? In other words, if I had to sacrifice one of these parameters, which would hurt my quality more? Lowering my bitrate to make a smaller file, or use a faster search to lower my encoding time?

I usually don't make long videos, but you never know. I just finished a 45 minute dance recital. But there is the possibility of longer videos in the future.

Thanks.

[hiptune]

I would'nt give up anyhting in quality if other people are going to view these videos.

For a 45 min video, I would not even use VBR, but stick with CBR of 8200kbs for starters.

Then you can then use some settings to gain speed if you really need faster encoding. But if this is something good, and others are going to view it, and your reputation is what they will see, stick to slowest encodeing. Go to bed, get up and author, and be proud of your work.

Jeff

[vhelp]

I agree too.. CBR all the way. Specially for your concert footage.

But wait. You said concert. Is that from home video cam you shot footage
of ??
If so, then that would mean you have an Interlace source. Definately
don't de-interlace this, and especially don't use any filters on it. You'll
only hurt the quality. Specially when you say lots of light flashes and other
strobing things, etc. With VBR, it's not going to cut it w/ all those
complicated scenes. IMO, encoding w/ a VBR is great for those scenes where
there are lots of them that are static (if I'm using the right word for it)
where there's hardly (if any) motion. There is where the bitrate will be
saved and used in other areas where needed. But, if your source is just too
busy (ie, sports too) CBR is what I would recommend. And, since your goal
is already headed for DVD, at 60 minutes, then using a bitrate of 9000 should
give you as best results as possible (depending on other factors)

If I misread anyone here, sorry.
Cheers,
-vhelp

[lordsmurf]

If everybody is TRULY concerned about quality, there are better encoders. Go buy PROCODER.

TMPGENC is honestly only good for encoding when you need advanced filters.

[n3vino]

Thanks to everyone for your replies. It was a Flamenco dance recital. It was an outdoors event held late afternoon where there was plenty of natural light. My wife is a Flamenco student and some of her classmates want copies. There may be more that want copies.

I shot it in widescreen with a Sony Digital 8 camera. I edited and created the AVI file with Mainactor. Then I encoded at 8000 cbr bitrate and at a slow high quality search speed (not the very slowest) with tmpgenc. Then I authored with tmpgenc and made six copies.

The quality turned out really good. I viewed it on an HD capable Widescreen TV.

The file size, including the Menu was 2.9 gig. So I had plenty of room on the DVD.

Thanks again.

[hiptune]

If everybody is TRULY concerned about quality, there are better encoders. Go buy PROCODER.

TMPGENC is honestly only good for encoding when you need advanced filters.

There you go again. Aother blanket statement. While I use Procoder sometimes, and tmpgenc others, at the higher bit rates, and at the slowest high quality/mastering settings, these two encoders are both exceptional. It is the lower bit rates (not used for 45 min. videos) where Procoder is said to shine above some of the others.

And now with dual layer discs gaining acceptance, I may never need lower bit rates below 7000kbp, so there is getting to be fewer reasons to spring for Procoder. The only major one is the choices of various quick time and windows media file encoding for web streaming. And one small feature is acceptance of Flash files. Other than that, the era of spending hundreds on a software encoder is as over as the reasons to spend $18.99 for a CD at Tower Records.

[lordsmurf]

I've never seen TMPGENC outperform PROCODER. Never.

[hiptune]

I did not say it out performed Procoder did I?

Please.

Jeff

[richdvd]

No comparison..Procoder over TMPGenc anyday.